Anagrafe della ricerca

ImPACT-AI - Development of inclusive quantitative photoacoustic imaging solutions enabled by ethical artificial intelligence

Durata:
24 mesi (2023 - 2025)
Responsabile scientifico:
Tipo di progetto:
PNRR – Missione 4
Ente finanziatore:
MINISTERO (MUR)
Codice identificativo progetto:
P2022C338Z
Ruolo PoliTo:
Partner

Abstract

Photoacoustics (PA) is a rapidly emerging imaging modality thanks to its inherent functional and metabolic capacity, sensitivity, depth penetration, non-invasiveness and radiation-free measurement of optical tissue properties. Hemoglobin is an active absorber of light and hence an ideal chromophore that enables quantitative PA-based analysis of vasculature in terms of morphology, network complexity and functional metabolism, i.e., blood oxygenation. Current limitations to PA imaging include artefact-ridden images when using traditional linear ultrasound probes (LUPs) for detection, or the need of ad-hoc detectors. Another challenge for PA imaging is that variations in skin tone, hence melanin content, can alter measurements between subjects in ways that the imaging device was not designed to anticipate. If ignored, this produces inaccurate functional information such as blood oxygenation levels based simply on the different skin color. ImPACT-AI aims to provide new PA imaging tools and artificial intelligence (AI) digital solutions for the inclusive quantitative imaging of vasculature in terms of complexity and functional oxygenation saturation analysis. Innovative methods for PA volume imaging using a LUP will be developed and innovative multi-angle illumination of the imaged area will be explored for high-resolution and artefact-reduced volumetric and functional PA imaging. The artificial intelligence methods will be developed by strictly following an “ethics by design” approach. AI methods, such as generative adversarial networks (GANs) and fully convolutional networks (FCNs), will be employed to optimize and enhance the acquired photoacoustic images. AI methods will be employed to automatically provide morphological and metabolic information about the acquired photoacoustic volumes. As a result of this project, new knowledge will be gained on how AI can enhance and enable photoacoustic imaging systems using a common LUP for detection and how an ethical and inclusive development process of AI techniques can compensate potential biases that could emerge due to skin tone variations. The project begins at an in-silico level designing, developing, and testing the new methodologies on simulated data. Then, the activity will advance towards in-vitro studies that consider phantoms. We will develop customized phantoms in compliance with the criteria defined by the International Photoacoustic Standardisation Consortium (IPASC). Our aim is the development of a reliable and low-cost platform that may even outclass the use of animals in terms of costs, repeatability, reproducibility, and serve as ground-truth to the AI solutions. The partnership of ImPACT-AI is a multidisciplinary team of experts in biomedical engineering, biophotonics, signal and image processing, and material science, thus providing an ideal synergy of background knowledge, tech know-how, case scenarios and outreach potential for the best scientific and socioeconomic impact.

Persone coinvolte

Dipartimenti coinvolti

Partner

  • C.N.R. - CONSIGLIO NAZIONALE DELLE RICERCHE
  • POLITECNICO DI BARI - Coordinatore
  • POLITECNICO DI TORINO

Parole chiave

Settori ERC

PE6_7 - Artificial intelligence, intelligent systems, multi agent systems
PE6_11 - Machine learning, statistical data processing and applications using signal processing (e.g. speech, image, video)

Obiettivi di Sviluppo Sostenibile (Sustainable Development Goals)

Obiettivo 3. Assicurare la salute e il benessere per tutti e per tutte le età

Budget

Costo totale progetto: € 240.000,00
Contributo totale progetto: € 240.000,00
Costo totale PoliTo: € 72.000,00
Contributo PoliTo: € 72.000,00